Ocean-MCP is an early-stage, actively-developed project. This policy reflects that: it's short, and it doesn't claim more process maturity than actually exists.

Please report suspected vulnerabilities privately using GitHub's private vulnerability reporting, rather than filing a public issue:

https://github.com/oceanmodeling/ocean-mcp/security/advisories/new

There is no dedicated security email for this project — use the link above. Include the affected server(s), a description of the issue, and reproduction steps if possible. This is a small project maintained without a formal SLA, but reports will be acknowledged and worked as soon as practical.

Ocean-MCP publishes 19 independently installable MCP servers, most of which are read-only clients over free public NOAA/ocean data APIs and require no credentials. The primary security-relevant surface is different:

• hpc-system-mcp and ufs-runner-mcp are the two servers that execute commands rather than just fetch data — they submit and monitor jobs on NOAA RDHPCS HPC systems (Slurm/PBS), and ufs-runner-mcp also stages files. Input handling, command construction, and any sandboxing or injection-hardening around these two servers is explicitly in scope and the area we'd most want a report about.
• The remaining servers are read-only data clients (CO-OPS, ERDDAP, NHC, Recon, STOFS, OFS, RTOFS, WW3, NDBC, USGS, winds, GOES, ADCIRC, SCHISM, and others) with no API keys and no write access to any upstream system. Vulnerabilities in how they parse or handle untrusted upstream API responses are still in scope, just lower severity than command execution paths.
• The publishing pipeline: all PyPI packages are published via GitHub OIDC trusted publishing (no long-lived tokens in the repo). Issues with the publish workflow or its permissions are in scope.

Every server in this repo is currently at a 0.1.x version. There is no long-term-support policy yet — only the latest published version of each server's PyPI package receives fixes. If you depend on one of these servers, pin a version and expect to need to upgrade rather than backport.

Please give us a reasonable window to investigate and, where applicable, publish a fix before any public disclosure. We'll coordinate timing with you on the advisory.